AI Top-of-Mind for Jan 22
Top-of-mind is the recent Microsoft Copilot Pro announcement. It combines priority access to GPT-4 and DALL-E 3, as well as ‘assist’ with Word, Excel, Outlook, and PowerPoint. $20/month, which is what I currently pay for just GPT-4. The free tier still is available. I’m seriously considering this swapping services.
On the security front, a good primer by Paul DelSignore on how to protect yourself against the growing sophistication of deepfakes. Does your family have a safeword? And maybe ‘Sports Illustrated’ deepfaked all of us with AI-created articles and profile photos. They are paying the price if you’ve looked at their publisher’s layoff announcement.
This also relates to recent coverage in the ‘NY Times’ as to how difficult it is to identify AI-generated images vs reality. Take the quiz and see how adept you are at telling the difference! I scored only 40%.
Turning to LLMs, if you’ve not yet seen it, a leaderboard of sorts maintained at ‘Huggingface.’
A few days back I wrote about the growing use by students of Gen AI and how teachers can counter it. Paul Pallaghy at ‘Medium’ offers some radical suggestions that may be worth trying out. No homework anyone?
Speaking of Gen AI, a not-so-great review by Anna Burgess Yang of Linkedin’s ‘Takeaways’ feature. From the posting:
LinkedIn’s Takeaways feature represents the worst type of use case for generative AI: replacing interesting content with uninteresting content. Removing the human element. It makes generative AI skeptics even more skeptical because they’ll roll their eyes and groan at the output.
Onto creative, where Moshe Sipper in ‘The Generator’ looks at ‘evolutionary algorithms’ for AI art creation vs the more common deep learning approach.
And continuing coverage of the Samsung Galaxy AI announcement. Lisa Eadicicco of CNET details her initial experiences with some of the Galaxy S24 Ultra’s more useful new features.
